RE: [PATCHv2 1/2] In the ARC assembler we have some special flag handling for the ".t" and ".nt" flags which only really applies to ARCv2, though currently the check is performed for all architectures.

[Claudiu Zissulescu <Claudiu dot Zissulescu at synopsys dot com>]

Hi,

This looks good from my side. Please wait for the Nick's (or any other binutils maintainer) approval. If you need help in committing the patch I can do that for you.

Cheers,
Claudiu

> -----Original Message-----
> From: Rinat Zelig [mailto:rinat@mellanox.com]
> Sent: Wednesday, March 15, 2017 10:13 AM
> To: binutils@sourceware.org
> Cc: claudiu.zissulescu@synopsys.com; Cupertino.Miranda@synopsys.com;
> Noam Camus <noamca@mellanox.com>; Andrew Burgess
> <andrew.burgess@embecosm.com>; Rinat Zelig <rinat@mellanox.com>
> Subject: FW: [PATCHv2 1/2] In the ARC assembler we have some special flag
> handling for the ".t" and ".nt" flags which only really applies to ARCv2, though
> currently the check is performed for all architectures.
> 
> Hi Claudiu,
> Thanks for the feedback.
> 
> Changes at second version are:
> 1. remove extra space after dot.
> 2. use ARC_OPCODE_ARCV2 macro.
> 
> Regards
> Rinat
> 
> In a later commit I'll be adding a new version of the ".nt" flag for an
> ARC700 extension (NPS400) which does not require this same special
> handling.
> 
> In this commit I have restricted the special flag handling to only apply if we
> are assembling for ARCv2.  This is a restructuring commit, and there should
> be no user visible changes after this commit.
> 
> gas/ChangeLog:
> 
> 	* config/tc-arc.c (assemble_insn): Only handle ".t" and ".nt"
> 	specially for ARCv2.
> 
> Signed-off-by: rinat zelig <rinat@mellanox.com>
> ---
>  gas/config/tc-arc.c |   18 +++++++++++++-----
>  1 files changed, 13 insertions(+), 5 deletions(-)
> 
> diff --git a/gas/config/tc-arc.c b/gas/config/tc-arc.c index a1fd71d..f806ca3
> 100644
> --- a/gas/config/tc-arc.c
> +++ b/gas/config/tc-arc.c
> @@ -4010,11 +4010,19 @@ assemble_insn (const struct arc_opcode
> *opcode,
>        if (!strcmp (flg_operand->name, "d"))
>  	has_delay_slot = TRUE;
> 
> -      /* There is an exceptional case when we cannot insert a flag
> -	 just as it is.  The .T flag must be handled in relation with
> -	 the relative address.  */
> -      if (!strcmp (flg_operand->name, "t")
> -	  || !strcmp (flg_operand->name, "nt"))
> +      /* There is an exceptional case when we cannot insert a flag just as
> +	 it is.  On ARCv2 the '.t' and '.nt' flags must be handled in
> +	 relation with the relative address.  Unfortunately, some of the
> +	 ARC700 extensions (NPS400) also have a '.nt' flag that should be
> +	 handled in the normal way.
> +
> +	 Flag operands don't have an architecture field, so we can't
> +	 directly validate that FLAG_OPERAND is valid for the current
> +	 architecture, what we do instead is just validate that we're
> +	 assembling for an ARCv2 architecture.  */
> +      if ((selected_cpu.flags & ARC_OPCODE_ARCV2)
> +	  && (!strcmp (flg_operand->name, "t")
> +	      || !strcmp (flg_operand->name, "nt")))
>  	{
>  	  unsigned bitYoperand = 0;
>  	  /* FIXME! move selection bbit/brcc in arc-opc.c.  */
> --
> 1.7.1